Kostas Bekris, Interesting problems in robotics and a few attempts to deal with them
This talk will provide an overview of the results and conjectures of my work as a graduate student at Rice, especially in relation to two fundamental problems in robotics: the localization ("Where am I?") and the motion planning problems ("How do I get there?").My first research project dealt with the problem of guiding a robot to a position without solving explicitly the localization problem and using only visual information. Inspired by this work on vision-based navigation, I also worked on vision-based localization, a challenging problem because a single camera is able to provide only bearings to environmental features and no distance information. An exciting collaborative project on localization was related to the problem of locating a wireless device using only the signal strength from the wireless Ethernet card. Finally, my doctoral thesis focuses on motion planning challenges, especially when the moving system has to obey dynamic motion constraints, such as an acceleration controlled car-like system that cannot move sideways. In my thesis, I had the opportunity to investigate faster techniques for solving such "kinodynamic" motion planning problems, investigate the challenges of the real-time operation of these algorithms and propose some solutions for distributed kinodynamic planning.